Explainable Machine Learning Methods for Ocean Worlds Mass Spectrometry Data: Biosignatures and Environmental Characterization

何をするか

Future astrobiological and geochemical investigations of ocean worlds (OWs) such as Europa and Enceladus will face challenges that can be addressed through science autonomy. While ML methods are potentially powerful tools for the prediction of geochemistry and biosignatures using IRMS data, many of these models are "black boxes" that reduce trust in predictions, and interpretable ML tools are needed to instill trust. In addition, methods are needed to diagnose false predictions for high-stakes predictions like extraterrestrial life.